Using Artificial Intelligence (AI) to Combine Visible Near Infrared (Vis-NIR) Spectroscopy With Lab-Based Techniques for Predicting Nitrogen Levels in Leaves to Enhance Fertilizer Application in Banana Farming

The current method for estimating fertilizer application for crops such as bananas involves removing a leaf or several leaves from the plant and sending them to a lab for chemical analysis. While this method is highly accurate, the process is slow and cost prohibitive for measuring abundant samples to make a well characterized evaluation of a farm. Using cloud based artificial intelligence (AI) to create and monitor predictive models for quantitative nitrogen could allow nonexperts to utilize in a practical way the power of Vis-NIR and predictive quantitation; A handheld field portable Vis-NIR instrument was used in combination with a cloud-based AI-based automated workflow along with initial lab assays to create near real-time predictions in the field for quantitative nitrogen; This study showed the RSQ was 0.9 and the MAE was 0.08 % nitrogen. The high value of the RSQ indicates an excellent correlation of the % nitrogen assay values and the spectral scans collected by the Vis-NIR instrument. The MAE was lower than the expected method error from the lab, which was 0.35 % with a 99 % confidence interval measure; These findings suggest that AIenabled quantitative modeling for % nitrogen prediction in banana leaves of the Cavendish variety is possible. Combined with a cloud-based mapping solution, banana growers could make well characterized evaluations of their crops.